What Happened
IREN (formerly Iris Energy) announced on May 1, 2026, that its Sweetwater 1 site in Texas has officially powered up its 1.4GW electrical system. This marks a key step in the company’s transformation from a Bitcoin mining company to an AI cloud computing provider.
Key Data
| Metric | Value | Meaning |
|---|---|---|
| Sweetwater 1 Power Capacity | 1.4 GW | Enough to power approximately 1.4 million households |
| Total Project Pipeline | 4.5+ GW | Total capacity under construction/planning |
| Deployed Nvidia GPUs | 150,000+ | Current online compute scale |
| Microsoft Contract | 5 years $9.7B | Secures long-term AI cloud service revenue |
Why It Matters
What Does 1.4GW Mean?
1.4 GW (1,400 megawatts) of power capacity, converted to data center scale:
- A standard AI data center rack consumes approximately 50-100 kW (GPU-dense deployment)
- 1.4GW can support approximately 14,000-28,000 racks
- Assuming 8 Nvidia GPUs per rack, the Sweetwater 1 site can theoretically house 112,000-224,000 GPUs
This aligns with IREN’s 150K+ deployed GPUs, confirming Sweetwater 1 carries its current main compute load.
Transformation from Mining to Cloud Provider
IREN’s predecessor was Iris Energy, a Bitcoin mining company. Under the AI wave, it completed one of the most successful transformations in the industry:
- 2024: Started transforming mining infrastructure into AI data centers
- 2025: Deployed first Nvidia GPUs, entered AI cloud services market
- 2026: Sweetwater 1 powered up, Microsoft $9.7B contract finalized
This timeline clearly shows a trend: Bitcoin mining companies’ existing power infrastructure and large-scale operational experience are exactly what AI data centers need most.
Microsoft’s $9.7 Billion Bet
The 5-year $9.7B contract means:
- Microsoft spends approximately $1.94 billion annually on IREN’s AI cloud services
- This does not include Azure’s own GPU procurement costs
- Microsoft is diversifying its supplier strategy to alleviate Nvidia GPU supply bottlenecks
The signal from this contract: hyperscale cloud providers are sparing no expense to secure AI compute supply, and IREN, with its power infrastructure and rapid deployment capabilities, has become a key supplier.
Landscape Assessment
Three Battlefields of the AI Infrastructure Race
| Battlefield | Core Capability | Representative Players |
|---|---|---|
| Chips | GPU/NPU design and manufacturing | Nvidia, AMD, Google TPU, custom chips |
| Power | Large-scale stable power supply | IREN, CoreWeave, traditional data center operators |
| Network | High-speed interconnect and data transmission | Cloud service providers, CDN companies |
IREN’s story shows that power is becoming the bottleneck and moat of the AI race. Whoever can obtain the largest scale of stable power at the lowest cost can lead in GPU deployment scale.
Implications for the Industry
- Mining company transformation is a shortcut: Bitcoin mining companies already have power infrastructure, land, and operational experience — the marginal cost of transforming to AI data centers is far lower than building from scratch
- Microsoft’s diversification strategy is accelerating: Beyond Azure’s own data centers, Microsoft is locking in independent suppliers’ compute through contracts
- AI compute “commoditization” is happening: When “non-traditional” players like IREN can provide AI cloud services, compute is shifting from a scarce resource to a purchasable commodity
Actionable Recommendations
| Your Role | Recommendation | Reason |
|---|---|---|
| AI startups | Watch emerging GPU cloud providers like IREN, CoreWeave | May have price advantages over AWS/GCP/Azure |
| Enterprise IT | Evaluate multi-supplier strategies, don’t lock into single cloud platform | Compute supply diversification helps reduce costs and supply risk |
| Investors | Focus on the intersection of power infrastructure and AI data centers | This is the most certain segment in the AI value chain |
| Developers | Test GPU instance performance and pricing across different cloud providers | Cost-performance differences may exceed expectations |
Sweetwater 1’s power-up is not an isolated event, but a marker that the AI infrastructure race has entered a new phase of “competing on power, scale, and speed.” As model capabilities become increasingly homogenized in 2026, compute supply capability may become the key variable in determining who wins.